Asynchronous Hydroclimatic Modeling for the Construction of Physically Based Streamflow Projections in a Context of Observation Scarcity

Abstract

Asynchronous hydroclimatic modeling is proposed for the construction of physically based streamflow projections over regions characterized by meteorological observation scarcity. The novel approach circumvents the requirement for meteorological observations by 1) calibrating quantile mapping transfer functions simultaneously to the parameters of the hydrologic model, 2) forcing the hydrologic model with post-processed climate simulations, and 3) intentionally ignoring the correlation between simulated streamflow values and observations. As a result, relative humidity, solar radiation and wind speed are integrated to a full hydroclimatic modeling chain, allowing the construction of streamflow projections forcing the Penman-Montheith reference evapotranspiration formulation over a forested catchment that flows into the St-Lawrence River, Canada. Results confirm a more accurate simulated hydrological response relative to a conventional hydroclimatic modeling chain employing reanalyses as description of the climate system. They also highlight the contribution to uncertainty in streamflow projections from biased climate variables issued by the reanalyses. The suggested framework assumes the hydrologic regime as a functional proxy to corresponding climate drivers. We believe the latter opens promising perspectives in the scope of producing more reliable estimations of waterrelated and energy-driven processes such as streamflow generation, snow accumulation and melt, river ice jams, water temperature, or vegetation growth under evolving climate conditions.

Producing hydrologic scenarios from raw climate model outputs using an asynchronous modelling framework

“…The application of statistical post-processed climate model outputs is criticized for three main reasons (e.g. Alfieri et al, 2015b, Chen et al, 2018Lee et al, 2018): (1) it disrupts the physical consistency between simulated climate variables; (2) it affects the trends in climate change signals imbedded within raw climate simulations; (3) it requires abundant good-quality meteorological observations, which are not available for many regions of the world, including some less common meteorological fields such as wind speed, relative humidity, and radiations (Ricard et al, 2020). More marginal critics raise the fact that statistical post-processing hides raw climate model outputs biases from end-users (Ehret et al, 2012), potentially blurring confidence attributed to resulting impact scenarios, and also, potentially misleading adaptation to climate change.…”

“…Chen et al (2017) quantified the hydrological impacts of climate change over North America, calibrating a lumped conceptual hydrologic model with raw RCM outputs over a recent past period. Ricard et al (2020) proposed to calibrate quantile-mapping transfer function concurrently to the parameter of a hydrologic model for meteorological fields for which observations are scarce or unavailable. Both studies operated a calibration with an objective-function that exclude the daily synchronicity of hydrologic events, such as the correspondence between observed and simulated empirical cumulative distribution functions (ecdfs), targeted quantiles, distribution moments, mean flows, or annual cycles.…”

“…Here, we propose a straightforward workflow, enabling the production of streamflow projections without post-processing climate model outputs and without using meteorological observations. The procedure is conducted inline with the asynchronous frameworks experimented by Chen et al (2017) and Ricard et al (2020). In essence, the proposed workflow translates raw climate model outputs into a corresponding simulated hydrologic response using an asynchronous modelling framework and to encrypt simulated hydrologic changes by defining perturbation factors for each streamflow quantiles.…”

“…The authors concluded that forcing a physically-based hydrologic model with regional climate simulations according to asynchronous modelling principles can improve the simulated hydrologic response over the historical period. Ricard et al (2020) implemented statistical post-processing of raw climate model outputs within the asynchronous modelling framework by calibrating quantile mapping transfer functions together with the parameters of the hydrologic model. Authors integrated relative humidity, solar radiation and wind speed to modelling chains and confirmed the improvement of the simulated hydrologic response in comparison to a conventional framework using reanalyses as a description of the reference climate.…”

“…of the hydrologic models. Instead, an asynchronous calibration loop(Ricard et al, 2020) is implemented forcing the hydrological model with raw climate model outputs over a recent past reference period, translating raw climate model outputs into a corresponding simulated hydrologic response. Since climate models cannot simulate the observed sequence of https://doi.org/10.5194/hess-2021-451 Preprint.…”